A startup whose product competes with GitHub Copilot and other AI-powered coding assistants has achieved unicorn status.
On Thursday, Codeium announced that it has closed a $150 million Series C funding round led by General Catalyst, valuing the corporate at $1.25 billion after the funding. The round, which also included participation from existing investors Kleiner Perkins and Greenoaks Capital, brings the corporate's total funding to almost 1 / 4 of a billion dollars ($243 million) – just three years after its founding.
Codeium's co-founder and CEO Varun Mohan told TechCrunch that Codeium hasn't even touched the $65 million Series B tranche it raised in January, just eight months ago when Codeium was valued at half a billion dollars.
“Although we have now barely touched our existing resources, we imagine this capital injection will enable us to significantly speed up R&D and growth while making even larger strategic investments,” he said.
Codeium was founded in 2021 by Mohan and his childhood friend and fellow MIT student Douglas Chen. Before Codeium, Chen was at Meta, where he helped develop software tools for VR headsets just like the Oculus Quest. Mohan was the engineering lead at autonomous delivery startup Nuro, liable for leading the autonomous infrastructure team.
The startup began as a radically different company called Exafunction, focused on GPU optimization and virtualization for AI workloads. But in 2022, Mohan and Chen saw an even bigger opportunity in generative coding and decided to rebrand—and refocus.
“Despite the influx of generative AI tools, developers still struggle with time-consuming coding tasks,” said Mohan. “Many of the AI-driven solutions provide generic code snippets that require significant manual work to integrate and secure into existing codebases. This is where our AI coding support is available in..“
Codeium's platform relies on generative AI models trained on public code and provides suggestions within the context of an app's entire codebase. It supports around 70 programming languages ​​and integrates with plenty of popular development environments, including Microsoft Visual Studio and JetBrains.
To lure developers away from Copilot and other competitors, Codeium initially released a generous free tier. The strategy appears to be working: Today, the startup has greater than 700,000 users and over 1,000 corporate customers, including Anduril, Zillow and Dell.
Quentin Clark, CEO of General Catalyst, said Codeium won a few of its larger contracts by taking a consistent customer-focused approach to product research.
“The team's approach has at all times been to follow its customers, which has driven the corporate to construct solutions on their terms — deployable in any environment and supporting more languages ​​than anyone else,” Clark said in a press release. “What Codeium has created will not be only a demo, an announcement or an idea — this can be a fully scalable business with large enterprises adopting the product across their entire organizations.”
Companies are sometimes cautious about revealing proprietary code to 3rd parties – for instance, Apple According to reports banned its employees from using Copilot last 12 months on account of concerns about confidential data leaks. To allay these fears, Codeium now offers a self-hosted installation option alongside its standard Software-as-a-Service plan.
Companies can now deploy Codeium's service on their very own hardware in the event that they so select, or they’ll select a hybrid configuration where they store their data on their very own devices and use Codeium's servers for his or her computing power.
Transferring data to the cloud at all times involves some risk, but Mohan claims that Codeium uses strong encryption. “We never train our proprietary generative autocomplete model with user data, never sell data, and make sure that all data transfers are encrypted,” he added.
Codeium has also taken steps to remove “non-permissively” licensed code (i.e., copyrighted code) from the datasets it uses to coach its AI models. Some code-generating tools trained with restrictively licensed or copyrighted code have been shown render that code in a certain way, which poses a liability risk (developers who include the code could possibly be sued). Mohan said that's not the case with Codeium, due to its approach to preparing and filtering training data.
“We also remove any residual data that resembles code that’s explicitly not freely licensed, just in case other people have copied code without providing proper attribution and licensing,” he added. “In addition, we have now state-of-the-art post-generation attribution filtering and logging in case these large probability models produce code that resembles public code, whether freely or non-freely licensed.”
But what about hallucinations? Most AI coding tools are notorious for making things up, which will be quite destructive in a company environment.
An evaluation by developer tooling startup GitClear found that generative AI tools have led to: more faulty code pushed into code bases lately. And a Purdue study found that over half of the answers given by OpenAI's ChatGPT to programming questions are incorrect. Security researchers have warned of the potential of such tools, amplify existing errors within the software.
A youngest A survey by cybersecurity firm Synk found that nine out of 10 developers are concerned in regards to the broader security implications of using AI coding platforms, but Mohan claimed that Codeium's supposedly superior, deeply contextual technology produces more trustworthy results than most.
“Our contextual awareness engine can base results on what already exists in a user's code base, leading to suggestions with fewer hallucinations and greater adherence to existing syntax, semantics and standards,” he said.
Whether benchmarks bear it out or not, Codeium's selling point appears to be resonating with the correct executives: Revenue hit eight figures this 12 months. Mohan said the 80-person San Jose-based startup plans to extend its headcount to 120 by 2025 to make an even bigger impression in a market stuffed with formidable competitors like Tabnine, Anysphere and Poolside.
Catching up with Copilot, the over 1.3 million paying users As of April, Codeium might be not on the cards – a minimum of not within the near future. Nor does it should be. As Mohan rightly noted, given the Widespread use of AI coding tools amongst developers (despite their reservations), even a small portion of this emerging segment is prone to be lucrative.
Polaris Research Projects that the AI ​​coding tools market can be value $27.17 billion by 2032.
“An excess of hype is a challenge for the industry,” Mohan said. “It will make it harder for any company to actually persuade end users that they’re on the forefront of what is feasible. But we imagine that truth-seeking and realistic AI firms like Codeium will eventually cut through that noise.”